Prediction of Electric Vehicle Battery State using SHAP Model

Abstract In this paper, we propose an explainable digital twin- based web platform for battery health prediction and detection in electric vehicles using machine learning and deep learning algorithms. The proposed system facilitates predictive maintenance by estimating the State of Health and detecting abnormalities through voltage, current, temperature, internal resistance, and cycle numbers. The system has two different users. The users are provided real-time predictions through a secure mechanism. The administrators are provided functionalities to manage the datasets and model training and assessment through dashboards. The proposed system has implemented Random Forest Classifier, Support Vector Machine Classifier, Decision Tree Classifier, and Long Short-Term Memory models. The performances have been calculated using R2 score and Mean Absolute Error. The proposed system indicates that the model performs better for the prediction of the time series battery health. The predictions have been explained using the SHapley Additive exPlanations method. The proposed web platform has been developed using the Django framework and has the capability to facilitate scalability and security for intelligent battery health. Keywords Digital Twin, Battery Health Prediction, Explainable Artificial Intelligence, Machine Learning, Deep Learning, Predictive Maintenance.